Why LLM features need their own threat list
On This Lesson 4 topics
This lesson is planned. The outline below lists the topics it will cover. The full lesson has not been published yet.
In this lesson
see what is new about LLM risks compared with ordinary web risks.
Planned topics
- A model cannot reliably tell instructions from data: everything in the prompt is text.
- Its output is unpredictable, so it must be treated as untrusted.
- The OWASP Top 10 for LLM Applications names main risks: prompt injection, sensitive information disclosure, improper output handling, excessive agency, system prompt leakage, unbounded consumption (verify the list).
- Old rules still apply: least privilege, validation, logging care.
- Threat-model the helper from JNR-056 the same way as the Shop API: assets, entry points, controls.
Draft outline details
Demonstration
a table: LLM risk / Shop helper example / control.
Check your understanding
Can the model tell your instructions from a customer's text?
Show answer
Not reliably; both are just text.
Practice
Apply this lesson with the module's lab, exercises and quiz.
Marking a lesson complete is optional and you can undo it any time. Your progress stays in this browser.